Velocity Reviews - Computer Hardware Reviews

Velocity Reviews > Newsgroups > Computing > Digital Photography > new 300D

Reply
Thread Tools

new 300D

 
 
David J. Littleboy
Guest
Posts: n/a
 
      08-24-2003

"Dave Martindale" <(E-Mail Removed)> wrote in message
news:bi99j3$b6e$(E-Mail Removed)...
> "David J. Littleboy" <(E-Mail Removed)> writes:
>
> >You've let yourself get sidetracked. The point with the SD9 is that any
> >"sharpness" (detail) it reports beyond what a 3MP Bayer camera would

report
> >is incorrect information that the camera has generated randomly.

>
> Not really. There are two principal differences between the flock of
> Bayer sensor cameras and the SD-9: Bayer interpolation vs. 3-colour
> sensing, and the antialiasing filter. I'm trying to discuss both.


But the 3-color silliness is easy to dismiss: it records information the eye
can't see, and is pointless. The human eye has lower color than luminance
acuity. If your print has adequate luminance resolution(a), the color
resolution will be adequate as well. If the luminance resolution is
inadequate(b), the color resolution will be similarly inadequate. But extra
color resolution in both cases would be meaningless, since you either
wouldn't see it (a) or wouldn't care (b) because the image is already soft.

> The point about the missing AA filter is your major criticism (and I
> agree), but Guido won't even admit there is a problem at all. On the
> other hand, he describes interpolation as something absolutely horrible,
> and I'm trying to point out that it's *not* as horrible as he makes it
> sound.


It's not horrible at all, it's not even a problem. The whole Foveon bit is a
complete joke since it solves a problem that doesn't exist. The 6MP DSLRs
are already doing better than 35mm for all but a very limited set of films.

> >The nasty thing about aliasing in digital systems is that you can't tell

the
> >difference between artifacts and real information; the artifacts look

good.
>
> Well, sometimes they look good. The additional high-frequency
> information that you get by downsampling using a box filter usually
> looks good.


Yes. That's why I downsample my MF scans<g>. (But I do need to do some more
experimenting, thanks for pointing the box filter point out: so far I've
just been using bicubic.)

> By the way, you shouldn't say "that the camera has generated randomly".
> Aliasing effects are not random at all - they're quite predictable given
> the original image and the characteristics of the camera.


But the artifacts are dependent on the exact relationship between the sensor
and the image and change with half-pixel displacements, and those
displacements are random. So the details of the artifacts are in practice
quite random.

> The problem
> is that (a) they're sometimes very ugly and (b) you can't invert the
> process (given an image with aliasing, you can't determine what the
> original looked like, and you can't remove the aliasing).


Exactly.

> >At best, the Foveon sensor might allow you to use a slightly less

aggressive
> >antialiasing filter, increasing your limiting resolution from 70% of

Nyquist
> >to 75% of Nyquist or so. But the truth of the matter is that even those
> >numbers are a bit overly optimimistic for digital imaging systems; using
> >less than three pixels to image a line is quite unreasonable.

>
> There's a tradeoff to be made here. The lower the cutoff frequency of
> the AA filter, the better the imaging system will behave (fewer
> aliasing artifacts) but the lower the resolution. Increase the cutoff
> frequency and you get more resolution but more errors. This applies to
> the demosaicing software as well. Looking at dpreview's resolution
> tests, you can see different manufacturers making different tradeoffs.
> In some cameras, the narrowing lines produce some wild colours before
> turning to grey. Others just go to grey quietly (the correct
> behaviour).


Yes. The Nikon D100 gets accused of being soft, but in the Japanese
magazines' test of artifacting, it does better than the 10D. I do think that
a Foveon sensor could provide 10% or so better resolution than Bayer, but
they talked themselves into a hole with the excessive hype and had to
provide twice (40% linearly) the resolution, and that it is not capable of.
The only real advantage of the Foveon sensor is for B&W landscapes, where
one would like to use a red filter, but that kills your resolution in a
Bayer camera. A lot of work for one trick.

> And then there's the SD9, which shows nice crisp bars where other
> cameras show grey. Unfortunately, it's the wrong number of bars with
> the wrong spacing between them - it's not what was on the test target
> in that location.


Exactly. Again<g>.

David J. Littleboy
Tokyo, Japan


 
Reply With Quote
 
 
 
 
Bart van der Wolf
Guest
Posts: n/a
 
      08-24-2003

"David J. Littleboy" <(E-Mail Removed)> wrote in message
news:bi9ek4$co0$(E-Mail Removed)...
SNIP
> Yes. That's why I downsample my MF scans<g>. (But I do need to do some

more
> experimenting, thanks for pointing the box filter point out: so far I've
> just been using bicubic.)


Using the Photoshop Filter|Pixelate|Mosaic function before (integer amount)
resizing may come close. A final Bicubic for getting the final bit of
resizing, and topping it of with a variation of USM often produces nice
results (depends on the subject as usual).

Bart


 
Reply With Quote
 
 
 
 
Michael Schnell
Guest
Posts: n/a
 
      08-24-2003
>The Bayer interpolated image contains two thirds of *false* information!

Strictly speaking this is correct. But It's a shortcoming of the (e.g. TIFF)
format the information is stored in (forcing 3 colored pixels). The RAW
format of a Bayer sensors camera avoids this problem.

-Michael

-Michael Schnell, Krefeld, Germany,
(E-Mail Removed)

 
Reply With Quote
 
Guido Vollbeding
Guest
Posts: n/a
 
      08-26-2003
Michael Schnell wrote:
>
> >The Bayer interpolated image contains two thirds of *false* information!

>
> Strictly speaking this is correct. But It's a shortcoming of the (e.g. TIFF)
> format the information is stored in (forcing 3 colored pixels). The RAW
> format of a Bayer sensors camera avoids this problem.


No, the shortcoming is in the crippled-color Bayer format,
not in the true-color image format.

Regards
Guido
 
Reply With Quote
 
Israel Rodriguez
Guest
Posts: n/a
 
      09-01-2003
On Sat, 23 Aug 2003 00:07:22 GMT, Todd Walker <(E-Mail Removed)>
with the help of a thousand monkeys banging on keyboards, was finally
able to type out the following:

>In article <(E-Mail Removed)>, http://www.velocityreviews.com/forums/(E-Mail Removed) says...
>> See my arguments about noise earlier in this thread.
>> This is a camera which records images with complete *real* picture
>> data, and thus noise, *not* artificial computer-generated images
>> modulated by one third of real picture data.
>>
>> Regards
>> Guido
>>

>
>That's it -- now I know you are a troll. Nobody is stupid enough to make
>a statement like that and be serious.


The name "Guido" wasn't a clue?

Israel Rodriguez

http://home.earthlink.net/~isrodriguez7/


 
Reply With Quote
 
 
 
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
EOS 300D & EOS 300D Rebel fatboybrando Digital Photography 2 03-26-2005 11:55 AM
Hong Kong prices on 300D, Canon S40+S45 + equipment for the 300D Hans Joergensen Digital Photography 3 01-26-2004 06:45 PM
BRAND NEW DIGITAL REBEL 300D FOR $789 Darius Alexander Digital Photography 10 09-27-2003 05:49 PM
Need wide angle lens for my new canon digital rebel 300D jahg Digital Photography 2 09-25-2003 03:30 AM
New Canon 300D vs. Pentax *ist-D jriegle Digital Photography 1 08-22-2003 08:16 AM



Advertisments